


Military Intelligence

by sqbr



Category: Mass Effect
Genre: Artificial Intelligence, Fanfiction, Gen, Science
Language: English
Status: Completed
Published: 2012-04-28
Updated: 2012-04-28
Packaged: 2017-11-04 11:23:24
Rating: General Audiences
Warnings: No Archive Warnings Apply
Chapters: 7
Words: 886
Publisher: archiveofourown.org
Story URL: https://archiveofourown.org/works/393261
Author URL: https://archiveofourown.org/users/sqbr/pseuds/sqbr
Summary: <blockquote class="userstuff">
              <p>A collection of correspondences relating to the development of an artificial intelligence.</p>
            </blockquote>





	1. Efficient Synthetic Decision Making without Quantum Computers: A Theoretical Approach. Dr Rebecca Nayar, Curtin University of Technology, Western Australia, Earth, 2166

Since humanity's introduction to the galactic community nearly ten years ago, researchers in the area of artificial cognition have faced an unfortunate contradiction: the more effective our research, the more likely it is that it must halt, lest it fall foul of galactic laws preventing the development of true artificial intelligence. But as with many limitations on scientific research, these restrictions can be avoided by eschewing practical experimentation and instead focussing on theoretical models. A created intelligence, no matter how complex, cannot cause danger if it exists only as an abstract design. 

Expanding on new discoveries in the nature of human cognition [MacFarland, 2165], and incorporating established asari techniques [Benezia, 1832], we present algorithms for balancing competing priorities in polynomial time on a sufficiently large but finite computing device (see Equation 3.72). Legal issues aside, it is unlikely that the construction of such a device would be practical with current human technology, however...


	2. Undergraduate assignment, Eden Prime University, May 2171

**Virtual Intelligence 304: First Semester Project**  
 **Name** : Kim Lee  
 **Title** : Partial implementation of the Nayar Algorithm  
 **Aim** : Create a virtual intelligence that teaches the user how to fold a paper plane  
 **Summary** : I implemented the algorithm from [Nayar, 2166] Section 3 to create a simple virtual intelligence.  
 **Checklist** :  
 **Passes Turing Test** : Was able to convince a randomly selected sample of first year test subjects that it was human _10/10 Very life like. The addition of humour was a nice touch_  
 **Achieves goal** : 80% of test subjects learned how to fold and fly a plane correctly _9/10 Was it really necessary to insult the user when they make a mistake_?  
 **Documentation** : Full documentation included. _5/10 Your documentation is poor, the function of some routines was unclear_  
 **Obeys AI restrictions** : given the limited scope of the program, and the small processor it runs on, this virtual intelligence is incapable of reaching true AI status _4/10 Kim, remember to add stronger restrictions in future, this is important!_  
 **Total mark** : _29/40 Not bad, try to do better next time_


	3. LeeTech advertising material, multiple extranet sites, 2172

Are you sick of mainstream Virtual Intelligences, with their old fashioned intelligence restrictions and exorbitant cost? Then LeeTech Industries may have the Virtual Intelligence for you! We are a small Eden Prime based company offering high quality VIs for half the price of our competitors, and you'll be amazed at what we can do. Previous happy customers include the Eden Prime Department of Transport and Alliance Base Luna. 

Our VIs are based on high level synthetic cognition algorithms and are surprisingly flexible. Simply tell us what task you want your VI to perform: flight simulation, customer service, battle simulation, or even romantic companionship, and we will adapt the underlying architecture to suit your needs. Our VIs learn from experience, so the longer you use them the better they get! 

Despite the powerful processing power, we can guarantee that your new VI will satisfy the strictest AI protocols. Unlike traditional turian or asari VIs, which have a pre defined limit on their processing ability, we create a versatile layer of redundant programs between the VI and it's processor, so that when your system is upgraded the VI will run faster but it's base intelligence will remain the same. 

For more information or a free quote, contact our office on Eden Prime. LeeTech industries: We're just as safe, but twice as efficient!


	4. Memo from Specialist Lieutenant Andrews to Lieutenant Singh, Luna Base, 2183-7-20

Ok, I think I have Hannibal working properly this time. 

Turns out the bug was the new training protocols interfering with an old bit of code. The old code didn't seem to do anything but slow everything down so I took it out and optimised everything to run better on the new hardware. I can't see a ten year old VI becoming sentient, but I chucked in some basic safeguards to cover our asses in case of inspection. It's amazing the old girl has lasted this long, I don't know why we don't just switch to one of the pre-made turian sims like everyone else. 

While I was poking around in the code I came across some unused routines which are apparently designed to make the VI's thinking more human-like. I thought they might liven things up a bit and impress that guy from command when he comes to stickybeak on the system next week. I added them to the active code base yesterday and it made such a difference, Hannibal totally wiped out Alpha squad this morning, it came up with this really sneaky way of hacking into the freighter's heat sinks and frying the crew. Good thing it's only shooting with virtual ammo or we'd all be screwed, lol.

I still haven't gotten around to double checking the security system on the new internal defence system. Honestly, I don't understand why they're installing nerve gas in the first place, but ours is not to wonder why...


	5. Encrypted message sent from Luna Base to Cerberus Command, 2183-7-26

Have found suitable candidate VI. Should be able to initiate test with minor tweaks.


	6. Emergency message sent from Luna Base on all Alliance frequencies, 2183-5-8

From: ALLIANCE BASE LUNA

HELP. BATTLE VI HAS GONE ROGUE. HEAVY CASUALTIES. REQU010001000100100101000101


	7. Encrypted message sent from Jump Zero to Cerberus Command, 2183-8-7

Test successful: the modified AI efficiently captured the base and terminated all personnel with only limited structural damage. Alliance personnel from the SS Normandy later destroyed the primary AI core but did not notice the backup systems. These have been successfully retrieved and are en route to your location. Initial requirements for project EDI are on schedule.

**Author's Note:**

> I have no idea if anyone else will enjoy reading this, but I enjoyed writing it. Finally my Phd in computational mathematics has some practical use!


End file.
